In recent years, image data readers such as optical character readers (OCRs) have been developed as devices for reading character information as image data for character recognition, and image data readers are now widely used in various industries to attempt to perform jobs efficiently.
For example, operators at windows of financial institutions attempt to perform their jobs more efficiently by processing documents more efficiently using the image data readers.
In particular, in order to increase efficiency in performing jobs such as processing the documents, it is required not only to process a large number of documents of the same type but also to automatically process a large amount of documents of various formats.
To cope with this, document processing devices provided with an image data reader are used. The image data reader of the document processing device for reading image data on a document is adapted to read image data based on a control from an electronic computer. In addition, for example, image scanners and facsimile machines are used as the image data reader. Furthermore, the image data reader can be an image data read and recognition device which can both read image data and recognize characters.
In addition, the electronic computer functioning as a controller for controlling the image data reader is constituted by an input means which is a keyboard or a mouse for inputting instructions from the operator and data, a computer main body and a display for displaying data or control information. In addition, recognition of image data read by the image data reader is performed by the electronic computer main body.
Furthermore, the document processing device is provided with a hard disk, which is connected to the electronic computer and stores in advance position information of character data to be recognized for each type of document and information designating types and numbers of characters (hereinafter, referred to as “definition information”).
Next, an operation will be described which is to be performed when the document processing device is used.
In recognizing character data described on, for example, an “ELECTRIC RATE BILL” using the image data reader, firstly, the operator operates the keyboard and designates definition information B corresponding to the type of a document (in this case, the document is regarded as an electric rate bill (Document B)).
Following this, at the electronic computer, the hard disk is accessed to draw the designated definition information B for the document, and the image data reader is notified of the information.
As this occurs, at the image data reader, reading of image data and recognition of characters can be performed, based on the definition information B which is a control information notified from the computer.
However, in a method for processing documents, as definition information is designated for each document to be read through the designation by the operator, the operator has to bear an additional work load, and there may occur designation errors with an increase in the number of pieces of definition information. Moreover, in a case where the operator is required to process thousands of types of documents, the designation by the operator is practically impossible.
To cope with this, there has been proposed a method for automatically reading documents without any designation by the operator, as described above, by describing in advance an specific ID number at a predetermined position on each type of document for discriminating the document of the type from others.
According to this method, in reading the image data of the document by the image data reader, the character recognition is made possible by firstly identifying the ID number put at the predetermined position and then using definition information (in this case, B) corresponding to the ID number.
However, in case the position on an optical reader portion of the image data reader where a document is set is changed in reading image data, for example, even in case document identical to the document whose definition information has already been stored in the hard disk, since a coordinate from the reference point (a physical origin) for an image such as a character data area and a graphic area does not match the coordinate in the definition information, it is determined that they are not in the same layout.
At the document processing device, there may be a case in which the character recognition process is not performed properly in case the character recognition of image data is implemented after the layout of the read image data has been made to match the layout of the definition information. Due to this, firstly, the reference points for the respective images are extracted, the coordinates of the respective images from the reference points are compared to each other, and whether or not the layouts match each other is determined.
Here, there are the following two methods for extracting a reference point for image data. As preconditions, a document to be read is a pre-printed document, and the printed position of the document on the form is controlled with high accuracy.
In a first method, in a case where image data of a document is read with an image data reader which can discriminate an end face of the document to be read from the background of the form to be read, for example, an upper left-hand side corner portion at the end face of the form is regarded as the reference point.
In addition, in a second method, in a case where image data of a document is read with an image scanner or a facsimile machine, since the read background cannot be discriminated from the end face of the form, a reference mark is printed in advance on a document form to be read, and this reference mark is extracted from the image data to become the reference point.
In this second method, since the reference point position is printed in advance as the reference mark, there is provided an advantage that even in case the position where a document is set on the reader is changed, the reference point can be extracted in a stable fashion.
However, in the document processing device, even in a case where the reference point is extracted accurately using the aforesaid method, since the character describing direction of image data does not become a correct direction in case the reading direction of image data is not the right direction, no character recognition process can be implemented.
For example, in case a money transfer request slip in which information is described horizontally is read from a wrong direction by the image data read and recognition device, the image data of the money transfer request slip so read is displayed on the display in the wrong direction.
Due to this, in recognizing characters by the document processing device, the operator determines whether or not the reading of the document is implemented from the right direction by looking at the display, and in case the reading direction of the document is determined to be wrong, the operator inputs from the keyboard a designation to rotate the image data of the document so read through 90 to 180 degrees, and a rotational correction process for the image data needs to be implemented so that the image data of the document can be displayed on the display in the right direction.
Incidentally, in executing jobs of transferring monies to accounts at banks, conventionally the operator inputs through the keyboard information on account number, names and amount of money which are described on a document which is a money transfer request slip, and as this occurs, the operator performs the input operation while watching the document and the display in an alternate fashion.
However, when the operator performs the input operation while watching the document and the display alternately, since the movement of the eyes occurs frequently, there exists a problem that the visibility is deteriorated and hence there may be caused an error in matching an item on the document with an item on the display.
To cope with this, there has been proposed a document processing device for reading a document with an image data read and recognition device and displaying the results of character recognition of the image data so read, as well as the image data itself on the display.
According to this document processing device, since information described on the document can be seen directly on the display, the movement of the eyes is reduced, whereby the occurrence of errors in matching between the character recognition results and the image data can be reduced when confirming and modifying the character recognition results.
On the other hand, in recent years, another method has become the main stream, in data processing jobs at financial institutions, in which a client-server system is used to do data processing collectively, at high speeds, at the server.
For example, it is proposed that clients, set at respective branches (sales points), and a server, set at a regional center of a financial institution, are connected via exclusive lines or public lines to form a client server system, so that document processing is implemented collectively by the server to thereby attempt to increase the efficiency of the document processing jobs.
In batch processing the documents by the server as described above, since the amount of data collected at the server becomes gigantic, it becomes impossible for the operator to do document processing at the server, and therefore, a system must be designed in which the server can automatically do document processing without the involvement of the operator.
To this end, the application of a document discriminating process using documents on which ID numbers are described to this client server system makes it possible for the server to automatically discriminate the types of documents and to read documents.
In addition, also at clients of this client server system, as has been described above, the results of character recognition of the image data read by the image data reader, as well as the image data itself can be displayed on the display, whereby the information described on the documents can be seen directly on the display, thus making it possible to reduce the occurrence of errors in matching the character recognition results and the image data when confirming and modifying the character recognition results.
However, in the aforesaid method for processing documents, only documents prepared exclusively for the document processing device can be read by the device, and general documents which have conventionally been in use cannot be used. Therefore, there exists a problem that special documents have to be prepared for use with such a document processing device.
Additionally, with the first reference point extracting method described above, while the print position of the document on the form needs to be controlled with high accuracy, in the case of a document printed with, for example, a word processor, papers are set manually, and the print position tends to vary every time printing is carried out, this causing the problem that using the upper left-hand side corner at the end face of the form as the reference point is not suitable for use with the word processor.
Furthermore, with the second reference point extracting method described above, the document to be read is the special document on which the reference mark is printed, this causing the problem that the reference point cannot be extracted with respect to general documents on which no reference mark is printed.
In addition, even with a method in which a special point on the front side of the document is used as the reference point, this method being adopted in the recognition technology used on document readers, there is also caused a problem that the method cannot function effectively if the layout of a document to be read cannot be identified to some extent.
Additionally, even in case a similar document as that which has been read is read again, there occurs a case in which an image which has been read does not match the image which was read before due to dust or a thin spot, and when this happens there may be a case where the same reference point cannot be extracted.
Furthermore, in order for the rotational correction processing to be automatically implemented at the server, the character recognition of the image data which is actually read is implemented, and whether the reading direction of the document is correct or not is determined by whether or not the character recognition is possible, this causing a problem that the increase in efficiency of the document processing job is remarkably reduced.
In addition, in the aforesaid method in which the character recognition results are confirmed and modified by displaying on the display the character recognition results, as well as the image data itself, since matching between the image data and the confirmation items is visually implemented, in case there are many items to be confirmed in the image data, there is also provided a problem that it is not possible to eliminate recognition errors in case where there are many items in the image data to be confirmed.
Furthermore, in a case where image data cannot be displayed on the display at one time, in order to refer to data disposed before, the display screen has to be scrolled down and, as this occurs by depressing a predetermined key with the finger the scroll operation, this results in a more complicated operation.